after the latest updates the boot is slowed down by nmbd.service that
receives
a timeout.
this happens in testing and fully updated unstable.
it only happens if "No local IPv4 non-loopback interfaces available".
to reproduce it just unplug the network cable and reboot or use a wifi
connection.

it works in bookworm:
====================================================================
root@bookworm:~# systemctl status nmbd
● nmbd.service - Samba NMB Daemon
Loaded: loaded (/lib/systemd/system/nmbd.service; enabled; preset:
enabled)
Active: activating (start) since Sat 2024-12-21 10:33:33 CET; 1min 25s ago
Docs: man:nmbd(8)
man:samba(7)
man:smb.conf(5)
Process: 635 ExecCondition=/usr/share/samba/is-configured nmb (code=exited,
status=0/SUCCESS)
Main PID: 721 (nmbd)
Status: "nmbd: No local IPv4 non-loopback interfaces available, waiting
for interface ..."
Tasks: 1 (limit: 2284)
Memory: 6.1M
CPU: 125ms
CGroup: /system.slice/nmbd.service
└─721 /usr/sbin/nmbd --foreground --no-process-group

dic 21 10:33:33 bookworm systemd[1]: Starting nmbd.service - Samba NMB
Daemon...

stop working in unstable and testing
====================================================================
root@unstable:~# systemctl status nmbd
× nmbd.service - Samba NMB Daemon
Loaded: loaded (/usr/lib/systemd/system/nmbd.service; enabled; preset:
enabled)
Active: failed (Result: timeout) since Sat 2024-12-21 10:54:34 CET; 48s
ago
Invocation: 609f91361bad477684c382f29ae33fe6
Docs: man:nmbd(8)
man:samba(7)
man:smb.conf(5)
Process: 722 ExecCondition=/usr/share/samba/is-configured nmb (code=exited,
status=0/SUCCESS)
Process: 992 ExecStart=/usr/sbin/nmbd --foreground --no-process-group
$NMBDOPTIONS (code=killed, signal=TERM)
Main PID: 992 (code=killed, signal=TERM)
Status: "nmbd: No local IPv4 non-loopback interfaces available, waiting
for interface ..."
Mem peak: 11.8M
CPU: 219ms

dic 21 10:53:04 unstable systemd[1]: Starting nmbd.service - Samba NMB
Daemon...
dic 21 10:54:34 unstable systemd[1]: nmbd.service: start operation timed
out.
Terminating.
dic 21 10:54:34 unstable systemd[1]: nmbd.service: Failed with result
'timeout'.
dic 21 10:54:34 unstable systemd[1]: Failed to start nmbd.service -
Samba NMB
Daemon.

21.12.2024 16:44, luca.pedrielli wrote:

That's an excellent reproducer, thanks!

Meanwhile, can you just disable netbios?  Are you sure you actually need it?
add 'disable netbios = yes' to the '[global]' section of samba.conf, and
it wont start anymore (or just disable the systemd service).

...

It looks like in samba 4.21 (as opposed to 4.17), this "waiting for interfaces"
comes before sending systemd notification.

Thanks,

/mjt

Dear Maintainer,

I have the same issue in SAMBA 4.22.1+dfsg-1

Regards

I remind both reporters of this bug, that I'm still waiting for the
answer to this question - does it work for you with no nmbd running?

netbios is obsolete and can't work in environments where no IPv4 network
is available.  To me it is easier to just drop nmbd from the default
config and be done with it, while for "proper" fix it might require
quite a lot of additional tweaking.

Thanks,

/mjt

Hello,

I ran into the same issue on Debian 13.
After installing samba the laptop would wait for the 1min30sec timeout for the nmbd service.
The laptop did not have a wired connection. It only had a wireless connection, which obviously wasn't active yet at the time nmbd got started.

I tried the wordaround you suggested and disabled netbios in the samba.conf file. This solved the issue for me.

I saw you were waiting for feedback, and figured I'd let you know. 🙂

Best regards,
Dimitri.

Thank you very much for the feedback.
I think I'll just switch netbios off by default now.

Thanks,

/mjt

Dear Maintainer,

With "disable netbios = yes" in smb.conf by default, they are no time out any
more.

Regards

Hi everyone,

I was also being affected by this.
I find nmbd useful to be able to just refer to the machines by their names
in controlled local networks, with a very simple setup. When I applied the
suggested workaround, I seemed to lose that functionality.

By inspecting the output of systemd-analyze critical-chain I found another
way - just disable the winbind service:

systemctl disable winbind

The boot time is thus much reduced and the name resolution offered by
nmbd continues
working as before. I see no degradation to any other function, so this
seems to be fine for my case.

Regards,